Microsoft Word - CoolCommPh2-Dr.GsFinalReport-COVERPG.docfor
Administration Building One
Dr. Lisa Gartland
The City of Tucson is committed to the promotion of sustainable
development practices and is considered a leader in the area of
energy efficiency. It has implemented numerous energy-saving
projects and won several awards for its efforts. The City has established a history
of promoting water and energy conservation, recycling, and participating in such
programs as Green Lights, Clean Cities and Climate Wise. In 1998, the mayor and
council requested City staff to develop policy options for sustainable design and
construction practices for City of Tucson facilities and in 1999, Tucson was
designated the Solar Capital of Arizona. The Creating Cool Communities project,
which addresses the increasing problem of urban heat islands, is still another City
effort to promote sustainability.
Urban Heat Islands are a recently noted phenomenon affecting large urban
centers globally. They occur when urbanized areas accumulate greater amounts of
heat than their rural surroundings. In cities throughout the world, on the hottest
summer day, the temperature is rising by up to one degree Fahrenheit every ten
years and can be six to eight degrees hotter than surrounding rural areas. Two main
factors are responsible for this phenomenon: 1) a decrease of trees and native
vegetation; and 2) dark, heat absorbing surfaces. This increase in temperatures
causes an increase in energy usage, energy costs, air pollution, greenhouse gasses
such as carbon dioxide and other indicators of global warming. The cities also
retain increased temperatures throughout the night, while the surrounding rural
areas cool off significantly. In 1992, the Cool Communities model was developed
to help mitigate the negative effects associated with urban heat islands.
In 1998, the City of Tucson applied for, and received, a planning grant from the
Public Technology, Inc.’s (PTI) Urban Consortium Energy Task Force for Phase
One of a two-phase project that addressed heat island mitigation. The goal of this
grant was to research and design a demonstration project that included all three
components of the Cool Communities model. The project incorporated the rooftop
of one office building and the two adjacent parking lots at the Thomas O. Price
Service Center in Tucson, Arizona. The 24,000 square foot building is part of a
larger City operations and maintenance complex. It was selected because of its
general set-up and ease of
CITY OF
TUCSON
PREFACE
monitoring due to its ongoing participation in the Department of Energy’s (DOE) Energy Star
Buildings program.
The intention of Phase Two of the Cool Communities project was to measure:
• improvements in comfort in and around the building;
• reductions in energy use due to the cooler roof surface and cooler surrounding parking lot
temperatures; and
• evaluate the cost effectiveness and durability of all roof, paving and landscaping materials
used in the project.
Due to lack of funds, Phase II was only partially completed; a highly reflective white coating
with reinforcing mesh was applied to the demonstration building’s existing roof. The existing
roof contains asbestos and had a temporary roof install seven years ago.
The following report by Dr. Lisa Gartland provides the results of monitoring building
temperature and energy usage of the demonstration building. The reductions in energy
consumption in the building due to topping the roof with a white reflective coating are quite
impressive. Nearly 50% energy usage reduction was recorded from one summer week, prior to
the coating, when compared to a week with similar weather conditions, after coating. The results
are even more impressive when taking into consideration that this building already had modern
energy efficient lighting, mechanical systems and building controls installed prior to baseline
monitoring. The existing roof also contains 2.5 inches of fiberboard insulation.
Although Dr. Gartland did not provide estimates of annual avoided energy consumption due to
the white topping, a reasonably accurate estimate can be calculated. Based upon the figures
presented in this report, as well as measured chilled water usage during the past summer, avoided
energy usage can be estimated at over 400 million Btu annually. This translates into an avoided
energy cost of nearly $4,000 annually. White topping the roof cost $24,993, therefore, the simple
payback period is just over six years. The rate of return is 16%. In addition to the monetary
savings the reflective roof top coating eliminated several roof leaks and extended the life of the
existing roof.
For more information, please contact the City’s Energy Manager, Vinnie Hunt, at (520) 791-5111
extension 311, or email him at VHunt1@ci.tucson.az.us. Vinnie can also be located at the City
of Tucson’s Thomas O. Price Service Center, 4004 South Park Avenue, Building 1, Tucson,
Arizona, 85714.
Cool Roof Energy Savings Evaluation for
Price Service Center Administration Building One
Building One of the City of Tucson’s Price Service Center has been monitored for weather
conditions, energy use and temperatures since the spring of 2000. This was done in anticipation
of the addition of a cool roof and cool parking lot features like lighter colored pavements and
added trees and vegetation to provide shade. In the first stage of this project, a white elastomeric
top coating was applied to Building One’s original metal and black surfaced roof between June
18 and June 29, 2001. This report evaluates the cooling energy savings resulting from this
rooftop change.
The energy savings are evaluated in two different ways. First, by looking directly at “raw” data
from a week before and a week after the cool roof was applied; and second, by adjusting this raw
data to account for weather and occupancy variations. This report recaps the roof construction
and the instrumentation used to monitor the building, then explains the data analysis work, and
finally gives roof surface temperature and energy savings results.
Table 1
Roof Surface Temperature Reductions and Cooling Energy Savings due to Cool Roofing
Raw Roof Temperature
Metal Roof 35 °F 46 °F
Average Temperatures - Black Roof 13 °F 17 °F
Metal Roof 13 °F 10 °F
Weekdays only for one summer week Raw Energy Use
(Btu)
May 7-13 & July 23-29 OFF at nite ON at nite
Hot roof cooling energy use 36,823,680 37,814,708 54,831,598
Cool roof cooling energy use 19,385,640 19,385,640 19,385,640
Savings 17,438,040 18,429,068 34,995,958
% savings 47.4% 48.7% 64.6%
Roof Construction & Building Instrumentation
The original roof on Building One is about 28,000 square feet in area. It is surfaced with a sturdy
aluminum foil over its low-slope areas (the vast majority of the roof area), and with a copper foil
that has turned black over more steeply sloped portions on the west side of the roof (furthest from
S. Park Avenue).
Slightly sloped, eastern
sloped section in the foreground
covered with a blackened copper foil.
Black surfaces have very low solar
reflectivities that make them hot.
Below the roof surface is a modified bitumen roof that was applied in 1993 as a repair to the
building’s original roof layer. The original layer consisted of a three-ply built-up roof surfaced
with an asphalt capsheet material, which was applied during the building’s construction in 1978.
Directly below the roof layers is a 2.5 inch thickness of fiberboard, which rests on a metal roof
deck. Below the roof deck is a partially enclosed air space about two feet high, containing
cooling ducts, wiring and other building equipment. Suspended ceiling tiles containing
fluorescent lighting fixtures are above the entire office space, with gaps between ceiling tiles
where structural members support the roof.
Underside of roof, air space, and ceiling tiles shown at a gap in the
ceiling around the roof support struts.
Table 2 lists the R-values of each of the roof and ceiling components for the original and new
cool roof installations. Note that an air space has a fairly constant R-value of about 0.04. Note
also that I’m assuming the fiberboard has an R-value-per-inch of 4.0, giving it an overall R-value
of 10. Addition of the new cool surface with embedded mesh has no effect on the R-value of the
entire roof assembly, so the R-value of the base roof and the new roof are the same.
Table 2. Calculation of roof & ceiling R values in hr-ft 2- F/Btu.
Component R-per-inch Thickness (in) Base R-value New R-value
Surface coating/mesh 0.1 0.020 --- 0.00
Foil surface 0.0 0.020 0.00 0.00
Modified bitumen 0.9 0.125 0.11 0.11
Built-up roof 0.9 0.125 0.11 0.11
Fiberboard 4.0 2.5 10.0 10.0
Metal roof deck 0.0 0.625 0.78 0.00
Air space 0.04 total --- 0.04 0.04
Acoustic tile 2.5 0.625 1.56 1.56
Total R-value 11.82 11.82
Monitoring equipment has been installed in and around the building. Table 3 lists the equipment
used, what it measures, the time interval between measurements, and when it was deployed, as
well as comments about data collection problems.
For the most part the data collection equipment worked smoothly. One important exception was
the rooftop pyranometer, which was reading erratically for about six months before this problem
was detected and fixed. This unfortunately meant that the solar radiation values during the
period before the roof coating were no good. Thankfully, these values could be reliably
estimated using a regression equation generated from summer 2000 data (see data analysis
section for details).
The only other significant problem with data collection equipment was the relatively late
installation of the chilled water BTU meter. This meter is obviously crucial to this analysis,
since it is measuring the cooling energy used by Building One. Since it wasn’t installed until
March 28, 2001, none of the data collected last summer or this winter could be used to analyze
energy savings.
Table 3. Monitoring equipment on Building One used to evaluate cooling energy savings.
Equipment Measurement Time interval Comments
Temperature
sensors
speed, humidity
Pyranometer Rooftop solar
December 2000 thru June 2001
Temperature &
for cooling
Data Analysis
Analysis Approach
To calculate the cooling energy savings due to the new cool roof, the data was analyzed two
different ways. First, data from before and after the roof was coated was directly compared. This
“raw” comparison was made carefully by picking before and after weeks with very similar
weather and building use conditions. Second, to account for even these small variations in
weather and building use, a regression equation was derived to describe the cooling energy use.
The “before” week chosen for comparison was Monday, May 7 through Sunday, May 13, 2001,
and the “after” week was Monday, July 23 through Sunday July 29, 2001. (Note that the roof
work was done between June 18 and 29, 2001.) These two weeks had the most similar weather
characteristics found during the weeks before and after the roof was coated.
Correcting “Before” Solar Radiation Data
Unidentified problems with the solar radiation data collection during the “before” period meant
that solar radiation values had to be regenerated. Figure 1 shows the actual “before” readings
during the week of May 7 through 13, along with the “after” week data. These problems with
extremely high readings and lots of data scatter began in December of 2000, and were not
identified and corrected until June of 2001, unfortunately not in time to get good summer 2001
data before the cool roof was installed.
Figure 1. Original solar radiation readings, showing problematic “before” readings
Past work on other roof monitoring projects has found that roof surface temperatures are highly
dependent on solar radiation values. Knowing this, it is possible to generate a correlation that
allows us to work backwards from roof surface temperatures to estimate the solar radiation
values during the May 7 through 13 “before” period.
Data from the summer of 2000, when the pyranometer was working correctly, was input to a
statistical regression analysis program to find the best correlation to predict solar radiation based
on rooftop surface temperature. Data from August 19 through 24, 2000, using only daytime
readings with non-zero pyranometer values, was regressed to find a multiple linear equation of
the form
Solar Radiation (V) = -1.158862 + 0.007134 Tblack + 0.008581 Tmetal, Equation. (1)
where Tblack is the surface temperature in degrees Fahrenheit measured on the black portion of
the roof and Tmetal is the surface temperature in degrees Fahrenheit measured on the metal
portion of the roof. This correlation has a robust R2 value of 0.805 with t-statistics of 19.3, 14.8
and 12.1 on the constant, Tblack and Tmetal coefficients, respectively.
When the correlation in Equation (1) is used to estimate/regenerate the pyranometer reading, it
produces some negative values instead of zero values when the sun is supposed to be down.
Therefore a further provision is made on Equation (1) that the value of solar radiation be set to
zero if Equation (1) produces a negative value. Actual and regressed solar radiation values are
plotted in Figure 2 for the period from August 19 through 24, 2000, to show how well the
correlation fits the actual data used to derive this regression equation. The corrected solar
radiation values for the week of May 7 through 13 are plotted in Figure 6.
Figure 2. Comparison of regression values and actual data for August 19-24, 2000
“Before” and “After” Weather and Building Data
Figures 3 through 8 compare the outdoor temperature, outdoor humidity, wind speed, solar
radiation, indoor temperature and building electrical use during the “before” and “after” weeks.
These are the weather and operating conditions that most strongly influence the cooling energy
use of the building.
Figure 3. Outdoor temperature before and after cool roof installation
Figure 4. Outdoor relative humidity before and after cool roof installation
Figure 5. Wind speed before and after cool roof installation
Figure 6. Solar radiation before and after cool roof installation
Figure 7. Indoor air temperatures before and after cool roof installation,
measured in an office cubicle
Figure 8. Building electrical use before and after cool roof installation
(does not include HVAC energy use).
Note that the building electrical use plotted in Figure 8 represents only the electrical energy
drawn by three electrical panels that serve the lights and outlets in the building. All cooling
energy use for the building was supplied via chilled water pipes running from a chiller in an
adjacent building.
Table 3 summarizes the average weather and building conditions during the “before” and “after”
weeks studied for this analysis.
Table 3. Average weather & building conditions before and after cool roof installation
Variable Before – May 7-13 After – July 23-29
Outdoor Temperature 82.7 °F 86.4 °F
Outdoor Relative Humidity 25% 45%
Wind Speed 2.4 mph 2.7 mph
Solar Radiation 300 V * (corrected) 272 V * (actual)
Indoor Temperature 77.6 °F 76.4 °F
Building Electrical Use 6165 kW 6139 kW
* Units of total volts read by the pyranometer, not converted into an energy flux.
The weather conditions in the “after” week are somewhat more severe than the “before”
conditions, with average outdoor temperature almost 4 °F hotter, 20% higher relative humidity,
and 0.3 mph faster wind speed. Indoor conditions are over 1 °F cooler after the cool roof is
added, while “after” building electrical use remains within 0.5% of its “before” value. More
severe weather conditions and cooler indoor temperatures during the “after” period mean that the
energy savings due to the cool roof will be under-predicted if based on a raw data analysis alone.
To get a true energy savings, unbiased by varying weather or building use conditions, it is
necessary to correct for these variations using correlations for roof surface temperature and
energy use.
“Before” Roof Surface Temperature Correlations
In order to eliminate variations due to weather conditions and get a true temperature reduction
due to the cool roof, a regression of surface temperature is derived. The “before” period data
from May 7 through 13 is statistically analyzed. The resulting multiple linear regression equation
for the surface temperature on the portion of the roof is:
Tmetal = 75.202 – 0.122556Tout + 58.761399Qsolar + 0.020878RH + 2.110335Wspeed,
Equation (2)
where Tmetal is in degrees Fahrenheit, Tout is the outdoor temperature in degrees Fahrenheit,
Qsolar is the solar radiation in units of Volts measured directly by the pyranometer, RH is the
outdoor relative humidity in percent, and Wspeed is the wind speed in miles per hour. The
overall R2 value for this correlation is 0.916, with t-statistics of 4.8, 0.7, 24.5, 0.2 and 4.2 on the
constant, Tout, Qsolar, RH and Wspeed coefficients, respectively. Usually variables with
t-statistics below 1.0 are dropped from a regression equation as insignificant variables. They are
kept here for consistency with the regression equation for the black roof surface (derived next),
where the outdoor temperature and relative humidity variables are significant.
The multiple linear regression for the surface temperature on the black portion of the roof found
from May 7 through 13 “before” data is:
Tblack = 18.125 + 0.509743Tout + 78.171259Qsolar +0.146907RH + 0.509349Wspeed,
Equation (3)
where Tblack is in degrees Fahrenheit and all other variables are in the same units as in Equation
(2). The R2 value of this correlation is 0.950, and t- statistics for the variables are 1.0, 2.6, 29.6,
1.3 and 0.9 for the constant, Tout, Qsolar, RH and Wspeed, respectively. Figure 9 shows how
well this regression correlation matches the actual “before” black and metal roof temperatures
measured during the week of May 7through 13.
Figure 9. Comparison of actual roof temperatures with regression results for
“before” data during May 7 through 13, 2001.
The correlations in Equations (2) and (3) are applied to the July 23 through 29 “after” data to find
true, roof surface temperatures that would have existed during this period if the original roof
were still in place.
“Before” Cooling Energy Use Correlation
The May 7 through 13 “before” data was statistically analyzed to produce a multiple regression
of the cooling energy use versus weather and building use variables. Since the cooling energy
use data was only collected hourly, the rest of the data was first condensed into hourly values by
either averaging 15 minute readings (most variables) or 30 minute readings (wind speed). The
weekend cooling energy use seemed to be on a different schedule from the weekday use, so the
regression analysis only used weekday data. The resulting equation to estimate “before” cooling
energy use characteristics is:
Cooling = 10705562 – 15115 Tout – 4756 RH – 157262 Qsolar + 28671 Wspeed Equation (4)
– 131724 Tin + 87828 Welec + 4960 Tmetal – 3048 Hour,
where Cooling is the cooling energy use in BTU, Tout is the outdoor temperature in degrees
Fahrenheit, RH is the outdoor relative humidity in percent, Qsolar is the solar radiation in Volts
measured by the pyranometer, Wspeed is the wind speed in miles per hour, Tin is the indoor
temperature in degrees Fahrenheit, Welec is the building electricity use in kW, Tmetal is the
surface temperature of the metal portion of the roof in degrees Fahrenheit, and Hour is an integer
from 1 to 24 representing the hour of the day. The R2 value of this correlation is 0.864, and the
t-statistics of each variable are 6.6, 3.7, 2.6, 1.5, 2.8, 8.5, 1.5, 2.4 and 1.4 for the constant, Tout,
RH, Qsolar, Wspeed, Tin, Welec, Tmetal and Hour, respectively. An additional assumption is
made that the cooling system is turned off between the hours of 7 pm and 6 am. Figure 10 shows
how well this regression correlation fits the actual “before” cooling energy use measured during
May 7 through 13. Note that the correlation does not do such a great job of predicting the
weekend energy use. This seems to be mostly due to this particular weekend’s cooling energy
use being on a different schedule than the weekday use, with the system turning on at around
noon on Saturday and staying on for the rest of the weekend.
Figure 10. Comparison of actual cooling energy use with regression results for
“before” data during May 7 through 13, 2001
Equation (4) is used later with the July 23 through 29 data to estimate the cooling energy the
original roof would have used under “after” weather and building conditions.
Roof Surface Temperature and Energy Savings Results
Raw Data Results
Figure 11 plots the actual measured roof temperatures during the “before” week of May 7
through 13, 2001 and the “after” week of July 23 through 29, 2001. Table 4 lists the peak and
average roof surface temperatures measured during these weeks. Without correcting for any
weather variations between these weeks, the surface temperatures show drops in temperatures of
70 °F and 35 °F at peak times and 13 °F averaged over the entire week.
Figure 11. Comparison of “raw” data roof surface temperatures measured during the
“before” week of May 7-13, 2001 and the “after” week of July 23-29, 2001
Table 4. Hot and cool roof raw data roof surface temperature comparison
Hot Roof Surface
May 7-13, 2001
Cool Roof Surface
July 23-29, 2001
Avg. Temperatures
Black Roof 98 °F 85 °F 13 °F
Metal Roof 98 °F 85 °F 13 °F
Figure 12 plots the raw data measurements of cooling energy use during the “before” week of
May 7 through 13, 2001 and the “after” week of July 23 through 29, 2001, and Table 5 lists the
actual cooling energy use and savings during these weeks. Without correcting for weather or
operational variations, the actual savings over the weekdays was 48%, and over the entire week it
was 61%.
Figure 12. Comparison of actual measured cooling energy use during the “before”
week of May 7-13, 2001 and the “after” week of July 23-39, 2001
Table 5. Cooling energy use & savings during “before” week of May 7-13, 2001
and the “after” week of July 23-29, 2001
Entire week Weekdays only
Savings 33,663,240 Btu 17,438,040 Btu
% savings 61.2% 47.4%
Weather-Corrected Results
Figure 13 plots the actual measured surface temperatures on the cool roof together with the
estimated hot roof temperatures based on the regression correlations of Equations (2) and (3) for
the “after” week of July 23 through 29, 2001. Both the black and metal portions of the roof
showed significantly lower surface temperatures during daylight hours. Table 6 summarizes the
peak and average temperatures during this week. The cool roof installation reduces the black
portion of the roof’s peak temperature by more than 70 degrees Fahrenheit, and the metal
portion’s temperature is reduced by over 40 degrees Fahrenheit.
Figure 13. Comparison of hot regressed & cool actual roof temperatures
during the “after” week of July 23-29, 2001
Table 6. Hot & cool roof temperature comparison during “after” week of July 23-29, 2001
Hot Roof Surface
Avg. Temperatures
Black Roof 102 °F 85 °F 17 °F
Metal Roof 95 °F 85 °F 10 °F
Figure 14 compares the actual cool roof cooling energy use to the energy use predicted by the
regression correlation of Equation (4) for the hot roof. Figure 14 sets the hot roof regression
correlation to zero between the hours of 7 pm and 6 am, assuming that the cooling system was
turned off at night.
The hot roof correlation of Equation (4) is not very accurate for weekend comparisons, so only
the weekdays are compared (Monday through Friday are the first five days plotted in Figure 14).
Table 7 lists the weekday cooling energy use and energy savings in the “after” week for
weekdays only. An impressive savings of 50% to 65% are realized during this week, assuming
the cooling system is turned off at night.
Figure 14. Comparison of hot roof & cool roof cooling energy use, assuming cooling is
turned OFF at night, during the “after” week of July 23-29, 2001
Table 7. Cool roof cooling energy use & savings during “after” week of July 23-29, 2001
Weekdays only OFF at night ON at night
Hot roof cooling energy use
(Regressed values)
(Actual values)
% savings 48.7% 64.6%
Effects of Insulation Versus the Cool Surface
One questions that is frequently asked is how effective is a cool roof compared to roof insulation.
The roof in this study already had a moderate R-11 level of insulation. How much energy would
a cool roof have saved without this insulation, or with higher insulation levels? Some quick
calculations are used to estimate these effects. The effective cooling load on the roof can be
estimated by using the equation:
Q = (Troof surface – Tinside) / R-value Equation (5)
The values of roof surface and inside temperatures from the “after” week of July 23 through 29,
2001, are used to estimate roof cooling loads with the actual ~R-11 insulation level and with
higher levels, all both with and without a cool surface. Table 8 shows these calculations at both
peak conditions and average conditions. Surface temperatures in Table 8 were estimated by
assuming that 1/8 of the total roof surface is covered black roof material and 7/8 is covered with
metallic surface material. At peak conditions the average roof temperature is then (198)(1/8) +
(165)(7/8) = 170 °F, and at average conditions the roof temperature is (102)(1/8) + (95)(7/8) = 96
°F.
From examination of Table 8, it is clear that adding a cool surface is on par with adding more
insulation in the roof. Further reductions in cooling load could have been made by adding more
insulation to the building, although an analysis should first be done to justify costs.
Table 8. Calculation of roof cooling loads and savings due to adding insulation
and a cool surface, for July 23 through 39, 2001 conditions
Surface
More insulation 170 76.4 19 / 30 4.9 / 3.1 42 / 64%
Cool surface 119 76.4 11 3.8 55%
Both 119 76.4 19 / 30 2.2 / 1.4 74 / 83%
Avg Conditions
More insulation 96 76.4 19 / 30 1.0 / 0.65 44 / 64%
Cool surface 85 76.4 11 0.78 57%
Both 85 76.4 19 / 30 0.45 / 0.29 75 / 84%
Conclusions
The cool roof installation has been a very successful energy savings measure, with savings of
between 50 and 65% of the building cooling energy depending on how the building is operated
after normal working hours. These savings are even more impressive when realizing that this
roof has a moderate level of insulation (around R-11).
It is clear that this building’s cooling energy use was previously dominated by the heat being
absorbed by the roof. The reasons for this can be found by examining the building. The roof
comprises 75 to 80% of the building’s surface area and is not shaded by any surrounding trees or
buildings. The outside walls and windows are protected from solar radiation by wide overhangs.
So the external cooling loads on this building are clearly dominated by the roof. Internal cooling
loads appear to have already been minimized with efficient lighting systems and good energy
management practices.
It is highly recommended that other buildings owned and operated by the City of Tucson also
have cool roofs installed during their next roof overhaul. Consideration should also be given to
the cost effectiveness of adding some type of insulation under the cool roof material (either a
spray polyurethane foam or some type of rigid insulation) to uninsulated roofs. This would
provide the combined benefits of reducing both the energy absorbed from the sun and the heat
conducted through the roof.